From tool idea to practical test: How we developed a migration support tool in one hour#

Is it possible to develop a functional tool in one hour with AI support? We tested it – and the result is now being evaluated for a website migration.

What was the challenge?#

We were faced with moving a website from one system to another.

The question was: Can this process be supported by a tool that analyses existing content, suggests improvements and exports it in various formats? But before investing weeks in development, we first wanted to know: Does the idea even work in practice?

What can the tool do?#

The tool processes entire website sections as structured data and improves the content through AI analysis. The process:

  • Import data: Structured export files are loaded into the tool (e.g. website sections with documents, images, downloads)
  • Analyse content: An AI language model (a programme that can understand and improve text) examines the structure and suggests optimisations (e.g. better headings, clearer text, removal of superfluous content)
  • Apply concept: The suggested improvements are applied to all content
  • Export results: Output in various formats (Markdown for further editing, ZIP archive for downloads, Word document for presentations)

The tool processes large amounts of data. A typical website section has 85 pages and is read in as a 12 MB file.

How did we proceed?#

  • Step 1 (20 minutes): We wrote a detailed description – what should the tool be able to do, how should it work technically, what are the requirements
  • Step 2 (1 hour): We handed this description over in its entirety to an AI language model, which developed code from it
  • Iteration (3 runs): First version → Add export formats → Refine interface details
  • Total effort: 20 minutes of planning + 1 hour of development
  • Result: 3 code modules, complete documentation, 4 export formats

We relied on familiar technologies: Python for the logic, a web interface framework for the user interface, container technology for the installation. This choice enabled rapid implementation.

Why did development proceed so quickly?#

Because we made the specification very detailed.

A clear description at the outset – both functional and technical – prevented errors and queries later on. We deliberately focused on the essentials. No feature proliferation, just what was really necessary. This resulted in focused code that is easier to maintain.

Key insights#

1. Specification beats iteration

The 20 minutes spent on the detailed description saved us several rounds of corrections. If you think carefully in advance about what the tool should be able to do, you’ll have less to fix later on.

The specification was not an abstract list of requirements, but rather concrete: Which functions exactly? Which technologies? Which output formats? This clarity is what made AI-supported development efficient in the first place.

2. Large amounts of data require planning

A website section with 85 pages becomes a 12 MB file with different content types. For such amounts of data, the AI language model needs a large context – in our case, 256,000 tokens (approximately 200,000 words).

This must be taken into account from the outset. Otherwise, the tool will work for small tests but fail when faced with real amounts of data.

3. From experiment to practical tool

The tool was actually only intended to test whether the idea would work. Now it is being evaluated by several teams and is helping with migration preparations.

This shows that if you can quickly turn ideas into working programmes, decisions can be made based on real tests – not on assumptions.

What can others learn from this?#

  • Invest time in the specification: The 20 minutes of planning were more important than additional development time
  • Just get started: Familiar technologies enable faster implementation than new, trendy frameworks
  • Keep it small: Only develop the necessary functions, no ‘nice-to-have’ features
  • Test on real data: Small examples always work – large amounts of data reveal the real challenges
  • Take prototypes seriously: What starts as a test can become a tool that others actually use

Conclusion#

✔ With clear specifications and AI support, you can develop working tools in 1 hour

✔ Rapid prototyping enables informed decisions based on real tests rather than assumptions

✔ Focusing on the essentials leads to maintainable code and usable results

This is part of a series on AI-supported development projects. The focus is on what can be learned from such experiments – not just on the results.

Backward Web-Helper Web Helper Short Forward